Oz Vision: Unlocking Colors Beyond Human Perception

Abstract

For displaying color images, we present Oz, a principle that involves directly controlling the activity of photoreceptors in the human eye through the delivery of light to individual cells. By exclusively activating M cone cells and avoiding the constraints imposed by cone spectral sensitivities, novel colors are theoretically possible. In practice, we confirm a partial colorspace expansion in that theoretical direction. By formal color matching on human subjects, it has been demonstrated that attempting to activate M cones exclusively elicits a color beyond the natural human gamut. The color, according to them, is a blue-green with unprecedented saturation. Oz colors are perceived by subjects in image and video form in additional experiments. Under fixational eye movement, the prototype delivers laser microdoses to thousands of spectrally classified cones. The programmable control of individual photoreceptors at the population scale is demonstrated by these findings as proof-of-principle.

INTRODUCTION

Oz is a novel color display principle that works by optically stimulating a population of photoreceptor cells on the retina to directly control their activation levels. Although reproducing the dynamic stimulation levels at each photoreceptor as imagery traverses the retina under eye movements requires exquisite precision, this cell-by-cell approach can theoretically display arbitrary colored visual imagery (see Fig. 1Creates an image viewer On a prototype Oz system that stimulates thousands of retinal cone cells, we conduct human subject experiments as proof of concept. Theoretically, Oz could enable the display of colors outside the well-known, limited color gamut of natural human vision (1). Because the M cone spectral response function lies between that of the L and S cones and overlaps completely with them, any light that stimulates an M cone cell must also stimulate its neighboring L and/or S cone cells in normal color vision (2, 3). However, Oz stimulation can, by definition, send a color signal to the brain that never occurs in natural vision because it only targets light to M cones and not L or S cones. Oz theoretically extends the human color spectrum to any L, M, or S color space (see Figure). 2Creates an image viewer The closest prior work for selectively exciting M cones is targeting light to only one (4–7) or two (8) cones at a time. In practice, we achieve a partial expansion of colorspace toward this theoretical maximum. Only visual pre-adaptation, such as bleaching L photopigment with red light prior to displaying green light, is able to selectively excite M cones in addition to cone-targeted methods (9, 10). However, due to their reliance on fleeting adaptation states and after-images, these perceptions are challenging to precisely measure (9, 11). Silent substitution (12, 13), a different approach, is capable of isolating activation changes in M cones; however, it necessitates baseline activation of the other cone classes and cannot display colors beyond the human gamut. Our Oz prototype, in contrast to these approaches, displays colors beyond the human gamut over a sufficiently large area for color matching, for extended periods of time, and within arbitrary colored imagery. Based on cone-targeted techniques (4–8) that make use of adaptive optics scanning light ophthalmoscopy (AOSLO), our Oz prototype is a proof-of-principle (14). The LMS type of 103 retinal cone cells (17) per subject is first spectrally preclassified using adaptive optics optical coherence tomography (AO-OCT) (15, 16). The Oz percepts are then produced by AOSLO by targeting 105 visible-wavelength laser microdoses per second to each cone cell and imaging the retina in infrared to nearly invisibly track eye motion at the cellular scale. The visual field of view of the prototype is a 0.9° square centered at 4° adjacent to a gaze-fixation target.

Formal color matching experiments were used to map the empirical colorspace coordinates of Oz colors in practice (Fig. 3Opens in a new window) and gathered qualitative hue and saturation evaluations. These tests showed that the prototype was able to successfully display a variety of colors in Oz, including orange, yellow, green, and blue-green with a 543-nm stimulating laser, which typically appears green. Additionally, color matching demonstrates that our attempt to stimulate only M cones produces a color that is outside of the normal human range. The ideal form of this new color, which we refer to as "olo," consists of only M activation. When compared to a neutral gray background, subjects report that olo in our prototype system appears blue-green with unprecedented saturation. In control experiments, Oz color matches "collapse" to the natural color of the laser, as expected, if we "jitter" the target location of each laser microdose so that it incorrectly lands on a random neighboring cell. Subjects discover that they must desaturate olo by adding white light before they can achieve a color match with the closest monochromatic light, which lies on the boundary of the gamut. Additionally, subjects clearly perceive Oz hues in the form of images and videos, such as a rotating red dot or an oriented red line on an olo background (Fig. 4Opens in an image viewer) and is unable to do so under the jitter control condition. Our prototype system must perform high-resolution retinal imaging, high-speed eye motion tracking, and low-latency stimulus delivery in order for Oz to perceive any color other than the stimulating laser's natural color (18). The perceptional sign that each of these system components is successfully working together is Oz's display of colors outside of the human gamut. An experimental platform for visual perception with a new class of precision, programmable control, and cellular scale is introduced by this technical achievement.